[LUM-745] Guard @Published property writes against no-op mutations to prevent SwiftUI attribute graph hang

[devin-ai-integration]

Fixes an app hang (Sentry VELLUM-ASSISTANT-MACOS-DH) caused by redundant @Published property writes in GatewayConnectionManager triggering unnecessary SwiftUI attribute graph invalidation. The @Published wrapper fires objectWillChange on every setter call regardless of value equality — when this happens while the attribute graph is already mid-update, it creates a re-entrant transaction that hangs the main thread for 2000ms+.

Guards all hot-path @Published writes with equality checks to suppress no-op mutations, extending the pattern already used by setConnected() (guard isConnected != connected) to assistantVersion, keyFingerprint, isConnecting, and isUpdateInProgress. Also removes a redundant else if branch in performHealthCheck() that wrote assistantVersion even when the value was unchanged.

Why this is safe: All changes are additive equality guards — they only suppress writes when the new value equals the current value. No state transitions or side effects are skipped.

Why not @Observable migration: GatewayConnectionManager has Combine subscribers outside SwiftUI ($isConnected.sink, consumers in ConversationManager, AppDelegate). A previous attempt to change this class's isolation model was reverted (PR #21711). The no-op guard approach fixes the root cause without architectural risk.

References:

• Apple: ObservableObject — documents that objectWillChange fires before any @Published mutation, with no equality check
• Sentry issue: VELLUM-ASSISTANT-MACOS-DH (event 7392246713)

---

Review & Testing Checklist for Human

• Build in Xcode locally — CI skips macOS/Swift checks, so a local build is required to confirm this compiles.
• Verify keyFingerprint guard doesn't affect credential invalidation: The side-effect block (if let oldFingerprint, oldFingerprint != newFingerprint) uses the independently-captured oldFingerprint local variable — confirm it still fires correctly with the guarded property write.
• Test: connect/disconnect cycle — verify connection status updates correctly in the UI after connecting and disconnecting.
• Test: health check loop — verify version and connection status still update when the assistant version actually changes.

Notes

• Same @Published cascade pattern as LUM-735 (MACOS-DE). The fix pattern is identical — guard writes with equality checks.
• The health check runs every 15s (or every 2s during updates), making assistantVersion and isConnected no-op writes the primary source of unnecessary invalidation.
• Non-Equatable published properties (latestMemoryStatus, latestModelInfo, permissionMode) are not guarded by this PR since they are written less frequently and would require adding Equatable conformance.

Link to Devin session: https://app.devin.ai/sessions/75c3a7510f2142029eb678ae055245c0
Requested by: @tkheyfets

---

[devin-ai-integration]

🤖 Devin AI Engineer

I'll be helping with this pull request! Here's what you should know:

✅ I will automatically:

• Address comments on this PR. Add '(aside)' to your comment to have me ignore it.
• Look at CI failures and help fix them

Note: I can only respond to comments from users who have write access to this repository.

⚙️ Control Options:

• Disable automatic comment and CI monitoring

[devin-ai-integration]

✅ Devin Review: No Issues Found

Devin Review analyzed this PR and found no potential bugs to report.

View in Devin Review to see 2 additional findings.

[tkheyfets]

@codex review

[chatgpt-codex-connector]

Codex Review: Didn't find any major issues. Another round soon, please!

ℹ️ About Codex in GitHub

Codex has been enabled to automatically review pull requests in this repo. Reviews are triggered when you

• Open a pull request for review
• Mark a draft as ready
• Comment "@codex review".

If Codex has suggestions, it will comment; otherwise it will react with 👍.

When you sign up for Codex through ChatGPT, Codex can also answer questions or update the PR, like "@codex address that feedback".